Shield TBM risk model based on Bayesian networks applicable in design stage

*Heeyoung Chung1), In-Mo Lee2) and Jeongjun Park3)

1), 2) School of Civil, Environmental and Architectural Engineering, Korea University,

Seoul 02841, Korea 3)

Advanced Infrastructure Research Team, Korea Railroad Research Institute, Gyeonggi 16105, Korea

3) jjpark@krri.re.kr

ABSTRACT

A risk management methodology that can be applied in the design stage of a shield tunnel boring machine (TBM) tunneling project is proposed in this paper. A Shield TBM Risk Analysis Model (STRAM) is developed based on Bayesian networks. The STRAM considers geological risk factors and TBM types, such as earth pressure balance (EPB) open mode, EPB closed mode, and slurry TBMs, and systematically identifies the potential risk events that may occur during tunnel construction. It can also quantitatively evaluate the degree of risk for the identified potential risk events by estimating the cost of counter-measures against event occurrence. The proposed methodology based on the STRAM can minimize the drawbacks of the TBM tunneling method, including the difficulty in substituting the machine type once it is selected and the excessive delay of the project caused by unexpected risk events. 1. INTRODUCTION During tunnel construction utilizing a shield tunnel boring machine (TBM), unexpected and severe risk events can occur, mainly due to the use of an unsuitable machine type for the specific ground conditions (Koh et al. 2010). Because it is impossible to substitute the machine type once it is selected, if risk events occur during tunnel construction, significant economic losses and/or delays of the project may occur. Even if the machine type is suitably selected, unexpected risk may still occur if the detailed specification of the TBM equipment is not suitable for the given ground condition (Kwak and Park 2009, Tóth et al. 2013). Therefore, selection of a suitable TBM type and detailed design of the machine are very important processes

1)

Graduate Student 2)

Professor 3)

Senior Researcher

that should be performed at the design stage, and a systematic and quantitative risk analysis that reflects the ground conditions for the project is essential. The risk register is one of the most commonly used risk management tools for TBM tunnel projects, but it is not based on a systematic approach to identify the potential risks and their causes. It is also impossible to analyze the risk quantitatively as the risk register does not include a risk matrix (Parsons 2008). Moreover, the risk register considers not only potential risk during construction, but also contractual, financial, and environmental factors, and thus, it is hard to build a specialized risk analysis for the construction process. The research on quantitative risk analysis for TBM tunnel construction is still ongoing. In this field, the study by Benardos and Kaliampakos (2004) on applicable risk analysis methods for the early stage of TBM projects is particularly notable. They identified rock parameters that will cause risks to tunnels and proposed a vulnerability index considering the uncertainty and variability of these parameters. These authors conducted a case study on the Athens Metro, and compared and verified their proposals with actual construction data. However, a comprehensive analysis on potential risks owing to risky ground conditions during TBM operation has not yet been conducted, and it is also uncertain how to deal with risky ground in the design stage. Sousa and Einstein (2012) provided a methodology using Bayesian networks to systematically assess and manage the risks associated with TBM tunnel construction. This methodology combines a geologic prediction model utilizing TBM data to predict the ground conditions ahead of a tunnel face and a construction strategy decision model that analyzes the risk of tunnel face collapse, and allows a choice between the open and closed modes of the earth pressure balance (EPB) shield TBM. However, this is a risk management system that is performed in construction rather than at design stage, and it only considers one risk event, the tunnel face collapse. Because the risk events that cause downtime during tunnel construction are diverse, it is crucial to analyze all of them to mitigate potential risks and delays. In this paper, we develop a Shield TBM Risk Analysis Model (STRAM) based on Bayesian networks to identify and quantitatively evaluate potential risk events that may occur during tunnel construction. It is based on the potential risk events, their main risk factors, and a series of risk-event occurrence processes that were identified by Park et al. (2016). A risk management methodology that is applicable to the shield TBM during the design stage is proposed using the developed model. The methodology can identify the potential risk events that may occur with each type of TBM, and can evaluate the degree of risk. The methodology is useful for selecting the optimized machine for a tunnel project that has the least risk. It also helps to develop a detailed machine design, able to counter-measure risk events that may occur when the selected equipment is used, because those events have been well identified. The risk management methodology proposed in this paper was applied to an actual tunnel project. Its validity was confirmed through the obtained suggestions on the selection of the optimized TBM type and assessments on the potential risk events that may occur during tunnel construction along the entire tunnel route.

2. BAYESIAN NETWORK THEORY Bayesian network theory is a causal network that utilizes available knowledge in an uncertain and complex situation to deduce a logical conclusion. It is a directed acyclic graph that consists of several nodes representing variables and directed edges connecting nodes to show a causal relation (Fig. 1). Each variable is finite, and has a mutually exclusive value. Each node has a set of probabilities (usually organized in an

n m table) that represents a conditional probability for the parent node (Jensen 2001). As shown in Fig. 1, node is called the “Parent” of node , and node is designated as the “Child” in a relation between nodes and . Except for nodes

and , which do not have any parent nodes, all the nodes have a conditional probability owing to connections with their parent nodes. The probability of each node

shown in Fig. 1 is represented by ( ), ( ), ( | ), ( | ), ( | ), ( | ), and ( | ), respectively.

Fig. 1 Schematic of a Bayesian network The Bayes' theorem that is used to deduce the Bayesian network is shown in Eq. (1).

( | ) ( | ) ( )

( ) ( )

Eq. (2) is established by adding a variable .

( | ) ( | ) ( | )

( | ) ( )

The joint probability of variables and is shown in Eq. (3), and conditioned on another variable we have Eq. (4).

( ) ( | ) ( ) ( )

( | ) ( | ) ( | ) ( )

The following relation can also be established when both variable and variable

have a conditional independence on variable .

( | ) ( | ) ( )

Eq. (5) shows that any information of variable cannot influence the probability of once the value of is revealed. By using this characteristic, if variables and

have a conditional independence on variable , Eq. (4) can be expressed as:

( | ) ( | ) ( | ) ( | ) ( | ) ( )

The joint probability of the Bayesian network ( ) in which * + can be obtained by Eq. (7), assuming the variables have a conditional independence among themselves, and by applying the chain rule.

( ) ∏ ( | ( ))

( )

where, ( ) are the parents of . For example, the joint probability of the Bayesian network shown in Fig. 1 can be represented as: ( ) ( ) ( ) ( ) ( | ) ( | ) ( | ) ( | ) ( | ) ( )

The Bayesian network can predict or deduce various probabilities by using the equations shown above, and it is especially useful to calculate the posterior probability when a series (network) of events has been formed. 3. IDENTIFYING POTENTIAL RISK EVENTS DURING SHIELD TBM OPERATION 3.1 Risk events and risk factors Comprehensive literature surveys and interviews with TBM experts were conducted to identify the risk events that could occur during tunnel construction utilizing a TBM, and to determine the main risk factors that could cause the events (Shirlaw et al. 2000, Kwak and Park 2009, Koh et al. 2010, Chong 2013, Tóth et al. 2013). As a result, the risk events were categorized into eight major risks as follows: cuttability reduction, collapse of tunnel face, ground surface settlement, ground surface upheaval, spurts of slurry on the ground, incapability of mucking, incapability of excavation, and water leakage. These risk events are shown in Table 1.

Table 1 Potential risk events during shield TBM construction

Top risk event Sub risk event Risk event

No. TBM type

Cuttability reduction

Excessive abrasion of cutters RE1-1

EPB and slurry TBM

Partial abrasion and damage of cutters RE1-2

Blockage of cutter RE1-3

Collapse of tunnel face RE2

Ground surface settlement RE3

Ground surface upheaval RE4 Slurry TBM Spurt of slurry on the ground RE5

Incapability of mucking

Large amount of ground water RE6-1

EPB TBM Breakdown of screw conveyer RE6-2

Blockage of screw conveyer RE6-3

Damage of pipe and pump RE6-4 Slurry TBM Blockage of slurry pipe RE6-5

Incapability of excavation

Machine jamming RE7-1 EPB and

slurry TBM

Insufficient torque and thrust force RE7-2

Misalignment/Off-route RE7-3

Water leakage RE8

The causes of these risk events are categorized into three types: geological factors, which exert the largest influence on risk derived from ground conditions along the tunneling route; design factors, derived from faulty TBM design; and construction management factors, caused by inappropriate construction management during TBM operation. Ten geological risk factors, nine design risk factors, and twelve construction management risk factors were identified, which are summarized in Table 2 (Park et al. 2016). Table 2 Classification of risk factors

Type Risk factor

No. Risk factor

A: GEOLOGICAL

FACTORS

G1 Hardness: hard or extremely hard rock, ground

containing large amounts of quartz

G2 Fractured zone or faults

G3 Weak ground

G4 Ground containing clay

G5 Ground with gravels and/or boulders

G6 Squeezing and swelling ground

G7 Mixed ground conditions

G8 Interface of different types of rock mass grade

G9 Ground with high water pressure

G10 Shallow cover depth

B: DESIGN

FACTORS

D1 Design of disc cutter and/or cutter bit

D2 Design of cutter head

D3 Design of opening ratio

D4 Design of torque and thrust force

D5 Design of segment

D6 Types of method for segment ring installation

D7 Selection of gaskets

D8 Types of method for backfill grouting

D9 Design of mucking facilities

C: CONSTRUCTION MANAGEMENT

FACTORS

M1 Delay of cutter replacement

M2 Delay of gauge cutter replacement

M3 Poor soil conditioning with foam and polymer

M4 Encounter of poor ground during EPB open mode

M5 Poor treatment of slurry

M6 Poor management of quality of mucking

M7 Poor management of amount of mucking

M8 Improper assembly of segments

M9 Poor management of backfill grouting

M10 Poor management of thrust force

M11 Improper reinforcement of borehole

M12 Poor management of tail seal

3.2 Causal relation between risk events and risk factors by using Bayesian networks The causal relation between risk events and risk factors proposed and summarized in Tables 1-2, was systematically analyzed by using a Bayesian network (Park et al. 2016). The results for the eight top risk events mentioned above are shown in Figs. 2-6. The details of the analysis are in Park et al. (2016).

G1 Hardness: hard or

extremely hard rock

Ground contains large

amounts of quartz

G4 Ground contains

clay

G5 Ground with

gravels and/or

boulders

G7 Mixed ground

conditions

G9 Ground with high

water pressure

Partial abrasion and

damage of the cutters

Sub Risk

Blockage of cutter

Sub Risk

Excessive abrasion of

the cutters

Sub Risk

M1 Delay of cutter

replacement

D1 Design of disc

cutter and/or cutter bit

Cuttability reduction

Top Risk

D2 Design of cutter

head

D3 Design of

opening ratio

Fig. 2 Causal relation between risk factors and cuttability reduction

Poor management of

face pressure

Sub Risk

Collapse of a tunnel

face

Top Risk

Ground surface

settlement

Top Risk

G5 Ground with

gravels and/or

boulders

G7 Mixed ground

conditions

G10 Shallow cover

depth

M3 Poor soil

conditioning with

foam and polymer

M7 Poor

management of

amount of mucking

G3 Weak ground

G8 Interface of

different type of rock

mass grade

M4 Encounter of

poor ground during

EPB open mode

D2 Design of cutter

head

D8 Type of methods

for backfill grout

G2 Fractured zone or

faults

Poor management of

face pressure

Sub Risk

Excessive face

pressure

Sub Risk

Collapse of a tunnel

face

Top Risk

G2 Fractured zone or

faults G3 Weak ground

G10 Shallow cover

depth

M11 Improper

reinforcement of

borehole

Ground surface

settlement

Top Risk

Ground surface

upheaval

Top Risk

Spurt of slurry on the

ground

Top Risk

Loss of slurry

Sub Risk

G7 Mixed ground

conditions

G8 Interface of

different type of rock

mass grade

G5 Ground with

gravels and/or

boulders

D2 Design of cutter

head

D8 Type of methods

for backfill grout

M5 Poor treatment

slurry

(a) EPB TBM (b) Slurry TBM

Fig. 3 Causal relation between risk factors and related to applying face pressure

G5 Ground with

gravels and/or

boulders

G9 Ground with high

water pressure

M6 Poor

management of

quality of mucking

Large amount of

ground water

Sub Risk

Incapability of

mucking

Top Risk

Breakdown of screw

conveyer

Sub Risk

G1 Hardness: hard or

extremely hard rock

Ground contains large

amounts of quartz

G4 Ground contains

clay

Breakdown of

conveyer belt

Sub Risk

M11 Improper

reinforcement of

borehole

G1 Hardness: hard or

extremely hard rock

Ground contains large

amounts of quartz

G4 Ground contains

clay

G5 Ground with

gravels and/or

boulders

G9 Ground with high

water pressure

Damage of the pipe

Sub Risk

Incapability of

mucking

Top Risk

Breakdown of the

pump

Sub Risk

Blockage in slurry

pipe

Sub Risk

D9 Design of

mucking facilities

(a) EPB TBM (b) Slurry TBM

Fig. 4 Causal relation between risk factors and incapability of mucking

G9 Ground with high

water pressure

Insufficient torque

and thrust force

Sub Risk

Misalignment/Off-

route

Sub Risk

Machine jamming

Sub Risk

Incapability of

excavation

Top Risk

Blockage of tail void

Sub Risk

G2 Fractured zone or

faults G3 Weak ground

G6 Squeezing and

swelling ground M2 Delay of gauge

cutter replacement

D4 Design of torque

and thrust force

Fig. 5 Causal relation between risk factors and incapability of excavation

G9 Ground with high

water pressure

Water leakage at the

segment lining

Sub Risk

Crack in segment

Sub Risk

Water leakage at the

TBM machine

Sub Risk

Water leakage

Top Risk

M8 Improper

assembly of segments

M9 Poor

management of

backfill grout

D7 Selection of

gaskets

D5 Design of

segment

M12 Poor

management of tail

seal

D6 Type of methods

for segment ring

installation

M10 Poor

management of thrust

force

Fig. 6 Causal relation between risk factors and water leakage 4. SHIELD TBM RISK ANALYSIS MODEL To conduct a risk analysis, it is necessary to identify risk factors, systematically propose potential risk events caused by these factors, and evaluate the degree of risk for the risk events. In this paper, a risk analysis model (STRAM) is proposed. The STRAM is based on the previously verified causal relation between risk factors and risk events, and is developed by using a Bayesian network that considers a conditional probability in order to evaluate the dependency of the causal relation. The schematic of the STRAM is shown in Fig. 7.

Counter-

Measure

Potential

Risk Event

Direct Cost

IndirectCost

Delay Level

TBM Type

TotalCost

GeologicalRisk Factor

Fig. 7 Schematic of STRAM

The rectangular shape shown in Fig. 7 is a decision node, which indicates that it does not have a probability. A chance node, represented by an ellipse, is defined as a chance node in the Bayesian network and has a conditional probability. A hexagonal shape is called a utility node, which denotes the cost of the parent node. In summary, the proposed model is classified into two parts: the causal relation between risk factors and risk events, and the quantitative evaluation of the potential risk events. 4.1 Causal relation between risk factors and risk events depending on TBM types As shown in Fig. 7, the TBM types and the geological risk factors are considered as the parent node (i.e., decision mode); potential risk events caused by these two parent nodes are designated as the child node. The TBM type was considered in our model because risk events and probability of their occurrences can vary depending on the TBM type, even under the same ground condition. Three types of shield TBMs are considered in this model: (i) EPB open mode, in which the machine is operated without any active face support; (ii) EPB closed mode, where the excavation chamber is completely filled with excavated muck; and (iii) slurry TBM, where the excavation chamber is completely filled with slurry. The EPB open mode is included in our study because it is used in practice if the rock condition ahead of the tunnel face is competent. In this case, the tunnel advance rate is superior to that in the EPB closed mode. However, the EPB open mode has a larger chance of risk event occurrence when encountering risky ground conditions. Based on its preferred usage and its higher potential risks under risky ground conditions, we included the EPB open mode in the proposed risk model to quantitatively evaluate the degree of such risks. The TBM type is categorized as a decision node, as its value does not have any probability. Of the various risk factors, only the geological risk factor was applied in the STRAM model. While design and construction management risk factors are also important, it is difficult to evaluate quantitatively the probability of occurrence of those risk events during the design stage. Design and construction management risk factors should generally be addressed as part of check lists that tunnel designers and/or workers at the job-site must follow to avoid them. On the contrary, because geological data can be obtained from the beginning of the project, geological risk factors can be objectively recognized and evaluated. This model considers ten geological risk factors, which are summarized in Table 2. Geological risk factors can be also categorized as a decision node. The potential risk events considered in the STRAM are summarized in Table 1. The causal relation between geological risk factors and risk events is derived using Bayesian network modelling and it is shown in Figs. 2-6. The risk events are grouped by either top risk or sub risk and assigned numbers (see Table 1), so that a counter-measure plan can be implemented for each risk event. The potential risk events (the child nodes shown in Fig. 7) are categorized as chance nodes, whose parent nodes are the TBM types and the geological factors. Then, we determine the probability of occurrence of each risk event, which is the conditional probability from the causal relation dependency between parent and child nodes. The probability can be obtained from a questionnaire administered to the tunneling experts.

4.2 Quantitative risk evaluation of risk events To evaluate quantitatively the risk events summarized in Table 1, the cost of counter-measures for restoring damages caused by risk event occurrences is estimated. In other words, the total expenditure, including the cost of construction materials, human resources, and additional costs owing to stoppage of construction should be considered. Therefore, the counter-measure considered in this risk evaluation method is not a proactive measure intended to prevent the risk event beforehand, but rather, a post-risk solution method that is to be utilized after the occurrence of a risk event. In order to establish a post-risk solution method (counter-measure) for each risk event, literature surveys as well as on-site interviews with tunneling experts were conducted. Counter-measure plans on the potential risk events are summarized in Table 3. Table 3 Counter-measure plan

Risk event Risk event No.

Counter-measure TBM type

Cuttability reduction

Excessive abrasion of cutters

RE1-1 Change of the disc cutter

EPB and

slurry TBM

Partial abrasion and damage of cutters

RE1-2 Change of the disc cutter

Blockage of cutter RE1-3 Removal of the blocking

material by laborer

Collapse of tunnel face RE2

Vertical grouting on the surface

Horizontal grouting in the tunnel

Ground surface settlement RE3 Vertical grouting on the

surface

Ground surface upheaval RE4 Vertical grouting on the

surface Slurry TBM

Spurt of slurry on the ground RE5 Removal of the slurry on the

surface

Incapability of mucking

Large amount of ground water

RE6-1 Discharge of the muck by

laborer

EPB TBM

Breakdown of screw conveyer

RE6-2 Repair of screw conveyor

Blockage of screw conveyer

RE6-3 Repair of screw conveyor

Damage of pipe and pump

RE6-4 Repair of pipe and pump Slurry TBM Blockage of slurry

pipe RE6-5 Repair of pipe

Incapability of

excavation

Machine jamming RE7-1 Excavation and ground

improvement near cutter head

EPB and

slurry TBM

Insufficient torque and thrust force

RE7-2 Installation of additional

equipment

Misalignment/Off-route

RE7-3 Re-alignment of TBM

Water leakage RE8 Counter-measure method for the water leakage at cutter head and back-fill grouting

The cost of a solution method (counter-measure) is divided into two parts, direct and indirect cost, and both are categorized as utility nodes (see Fig. 7). The direct cost includes the costs of re-design, removal and re-building, among other aspects, and it is largely job-site dependent. The indirect cost includes the costs of waiting time, construction management, and delay charges, and it is also job-site dependent. In order to estimate the indirect cost more easily, a chance node called "delay level" is adopted between the chance node of counter-measure and the utility node of indirect cost shown in Fig. 7. It considers the fact that the construction period to apply the counter-measure varies depending on in-situ conditions. Six delay levels are proposed, as documented in Table 4. If regardless of the counter-measuring action there is no construction delay, the delay level is 0. An increase in delay level indicates longer delays caused by counter-measuring actions. Table 4 Delay level

Delay level Construction delay description

0 No delay

1 Below 4 days

2 5 days – 14 days

3 15 days – 1 month

4 1 month – 6 months

5 More than 6 months

The degree of risk represented by an expected utility ( ) can be expressed as:

( ) ∑ ( | ) ( | )

* ( ) ( | ) ( )+ ( )

where, represents the degree of risk (i.e., cost) of the risk event, is a conditional probability for the relevant node, is a direct utility cost for the counter-measure, indicates an indirect utility cost, represents the TBM type, is the geological

risk factor, stands for potential risk event, is the counter-measure, and represents the delay level. The aforementioned conditional probability of ( | ) can be calculated based on the results of a questionnaire given to the TBM tunneling

experts. The conditional probability of ( | ) has a value of 1 if a risk event occurs, and a value of 0 in the absence of a risk event occurrence. ( | ) can also be obtained based on a questionnaire. The direct cost ( ) and the indirect cost ( ) are calculated once the counter-measure scheme is decided, which depends on the job-site condition. 4.3 Risk management methodology to be used in design stage The risk management methodology to be used in design stage can be established by applying the STRAM mentioned in the previous section. The purpose of creating the risk management methodology in design stage is to identify the dangerous locations along the tunneling route, recognize potential risk events in each risky location, and evaluate the degree of risks depending on the TBM type. Then, the best TBM type is chosen, defined as the machine that has the least degree of risk when assessing and summing the risk degrees for the entire route of the tunnel. The next step is to prepare the risk events that may occur during tunnel construction even with the choice of the optimized TBM. The process of the proposed risk management methodology to be used in design stage is summarized as the flowchart shown in Fig. 8.

Selection of geological risk factors

Risk event assessment using STRAM

Selection of counter-measure and assessment of its cost

Degree of risk assessment using STRAM

Choice of optimum TBM machine type

Presentation of potential risk events along tunnel routes

Fig. 8 Process of risk management methodology in design stage

5. A CASE STUDY The risk management methodology proposed in this paper was applied to an actual tunnel project, and its validity was confirmed through the derived suggestions for the selection of an optimized TBM type and the assessments on the potential risk events that may occur during tunnel construction along the whole tunnel route. 5.1 Overview of the project The project in which the risk management methodology proposed in this paper was applied is the first undersea tunnel in Korea with maximum water depth of 80 m. Its total length is 7.985 km, while the length of the tunnel is 6.297 km. It is currently under construction by the conventional tunneling method. It was a turnkey project and the mechanized tunneling method utilizing a shield TBM was proposed by another company, but was not awarded. The present risk analysis was performed assuming the project was constructed by a shield TBM. The proposed cross-section of the tunnel is shown in Fig. 9 with the tunnel diameter of 9.7 m.

Fig. 9 Shield TBM tunnel cross-section of the project 5.2 Geological risk factors The first step to apply the proposed risk management methodology is to identify geological risk factors for the job-site. The detailed geological profiles of this project were well documented by Suh et al. (2010). An overview of the longitudinal geological profiles is shown in Fig. 10.

Fig. 10 Longitudinal geological profile of the project The geological risk factors were taken from Suh et al. (2010). These risk factors include hard or extremely hard rock (G1), fractured zone or faults (G2), ground containing clay (G4), mixed ground conditions (G7), interface of different types of rock mass grade (G8), and ground with high water pressure (G9), which are summarized in Table 5. The locations of these geological risk factors along the entire tunnel route are summarized in Table 6. Table 5 Potential risk event and probability of risk event occurrence

Geological risk factor

Potential risk event TBM type Occurrence probability

G1

Hardness: hard or

extremely hard rock,

ground containing

large amounts of

quartz

RE1-1 Cuttability reduction

(Excessive abrasion of cutters)

EPB open 0.77

EPB closed 0.9

Slurry 0.77

RE6-2 Incapability of mucking (Breakdown of screw

conveyer)

EPB open 0.34

EPB closed 0.37

RE6-4 Incapability of mucking (Damage of pipe and

pump) Slurry 0.41

G2 Fractured zone or faults

RE2 Collapse of tunnel face

EPB open 0.72

EPB closed 0.4

Slurry 0.2

RE7-3 Incapability of excavation (Misalignment/Off-route)

EPB open 0.28

EPB closed 0.28

Slurry 0.31

G4 Ground

containing clay

RE1-3 Cuttability reduction (Blockage of cutter)

EPB open 0.34

EPB closed 0.53

Slurry 0.73

RE6-3 Incapability of mucking

(Blockage of screw conveyer)

EPB open 0.22

EPB closed 0.24

RE6-5 Incapability of mucking

(Blockage of slurry pipe) Slurry 0.31

G7 Mixed ground

conditions

RE1-2 Cuttability reduction (Partial abrasion and damage of cutters)

EPB open 0.53

EPB closed 0.6

Slurry 0.47

RE2 Collapse of tunnel face

EPB open 0.72

EPB closed 0.4

Slurry 0.2

G8

Interface of different types of

rock mass grade

RE2 Collapse of tunnel face

EPB open 0.72

EPB closed 0.4

Slurry 0.2

G9

Ground with high

water pressure

RE1-2 Cuttability reduction (Partial abrasion and damage of cutters)

EPB open 0.53

EPB closed 0.67

Slurry 0.54

RE6-1 Incapability of mucking

(Large amount of ground water)

EPB open 0.78

EPB closed 0.54

RE6-4 Incapability of mucking (Damage of pipe and

pump) Slurry 0.6

RE7-2 Incapability of excavation (Insufficient torque and

thrust force)

EPB open 0.28

EPB closed 0.28

Slurry 0.31

RE8 Water leakage

EPB open 0.47

EPB closed 0.34

Slurry 0.22

Table 6 Geological risk factors and degree of risk

Location Geological risk factor Degree of risk*

EPB open

EPB closed

Slurry

0+412~0+465 (53 m)

Mixed ground conditions 346 221 126

1+810~1+935 (125 m)

Mixed ground conditions 817 521 298

2+010~2+020 (10 m)

Fractured zone or faults Ground with high water pressure

473 373 295

2+122~2+137 (15 m)

Fractured zone or faults Ground with high water pressure

Ground containing clay 486 391 323

2+240~2+250 (10 m)

Fractured zone or faults Ground with high water pressure

Ground containing clay 486 391 323

2+593~2+613 (20 m)

Interface of different types of rock mass grade

112 62 31

2+603~2+618 (15 m)

Fractured zone or faults Ground with high water pressure

448 356 294

2+793~2+808 (15 m)

Fractured zone or faults Ground with high water pressure

Ground containing clay 456 368 311

2+835~2+845 (10 m)

Fractured zone or faults Ground with high water pressure

Ground containing clay 456 368 311

3+285~3+295 (10 m)

Fractured zone or faults Ground with high water pressure

448 356 294

3+485~3+495 (10 m)

Fractured zone or faults Ground with high water pressure

448 356 294

3+821~3+932 (111 m)

Hardness: hard or extremely hard rock, ground containing large amounts of quartz

291 329 230

4+694~4+908 (214 m)

Hardness: hard or extremely hard rock, ground containing large amounts of quartz

561 633 444

5+155~5+160 (5 m)

Fractured zone or faults Ground with high water pressure

448 356 294

6+360~6+370 (10 m)

Fractured zone or faults Ground containing clay

130 85 65

6+739~6+958 (219 m)

Mixed ground conditions 1,431 912 522

6+797~6+802 (5 m)

Fractured zone or faults 117 67 36

Total 7,951 6,145 4,491

*Unit: US dollars 5.3 Potential risk events and probability estimation The potential risk events caused by the geological risk factors mentioned above (G1 through G9) were identified and are summarized in Table 5. As shown in Table 5, some risk events such as ground surface settlement or upheaval and spurts of slurry on the ground were not considered in this case study, because the project is being constructed under deep sea. The conditional probability of each risk event (chance node) with its parent nodes (TBM type and geological risk factor) was then obtained. To

accomplish this, a questionnaire survey was administered to approximately 10 TBM tunneling experts with more than 20 years of experience. The results are also shown in Table 5. 5.4 Counter-measure plan and cost estimation For the identified risk events summarized in Table 5, the counter-measure plan summarized in Table 3 can also be adopted for the job-site. The cost of each counter-measure is estimated after a consultation with TBM experts along with cost-estimation specialists. The direct and indirect costs were estimated considering the delay level of each counter-measure applicable to the job-site. The details are in Park (2015). 5.5 TBM choice and potential risk events summary

The degree of risk, i.e., the utility value ( ), for each risk event was estimated using Eq. (9). Then, by adding the degree of risk of the corresponding events, the final degree of risk for each geological factor was calculated for the three machine types. The results are summarized in Table 7. The factor of ground with high water pressure has the highest degree of risk, as the solution method (counter-measure) may necessitate the installation of additional jacks to mitigate this risk event related to insufficient torque and thrust force under high water pressure conditions. Based on this, the final results of risk degree for the entire tunnel route are also shown in Table 6. These results suggest that the slurry TBM has the least degree of risk. The EPB open mode increases the risk by 177% compared to the slurry TBM. Similarly, even the EPB closed mode exhibits a 137% increase in risk compared to the slurry TBM. The slurry TBM is generally known as the best choice under high water pressure conditions, and the results of this study systematically confirm this. Table 7 Degree of risk for each geological risk factor

Geological risk factor

Degree of risk*

EPB open

EPB closed

Slurry

G1 Hardness: hard or extremely hard rock, ground

containing large amounts of quartz 52.39 59.19 41.47

G2 Fractured zone or faults 116.78 66.97 36.35

G4 Ground containing clay 13.40 18.43 28.39

G7 Mixed ground conditions 130.68 83.33 47.63

G8 Interface of different types of rock mass grade 112.07 62.26 31.13

G9 Ground with high water pressure 355.92 306.06 258.75

Ground with high water pressure (Not including additional equipment installation)

164.96 115.11 47.33

*Unit: US dollars Even if a slurry TBM is adopted for the project, there are still potential risk events in tunnel works, such as excessive abrasion of cutters, partial abrasion and damage of cutters, blockage of cutters, collapse of tunnel face, damage of pipe and pump, blockages in slurry pipe, insufficient torque and thrust force, misalignment/off-route, and

water leakage. Potential risk events that may occur in the locations with geological risk factors have to be considered when designing and manufacturing the selected TBM. 6. CONCLUSION In this paper, a risk management methodology for use in the design stage of a tunnel project utilizing a shield TBM is proposed by developing systematic and quantitative analyses for risk events that may occur during construction. The potential risk events during shield TBM construction are identified and classified into eight categories: cuttability reduction, collapse of tunnel face, ground surface settlement and upheaval, spurts of slurry on the ground, incapability of mucking and excavation, and water leakage. In addition, sixteen sub-categories of risk events were also identified and classified. The causes of risk events were classified into three categories: geological, design, and construction management risk factors. Ten geological, nine design, and twelve construction management risk factors were determined. Bayesian networks were used to establish the causal relation between risk factors and risk events, which depends on the TBM type. Based on the causal relation between geological risk factors and risk events, a risk analysis model, STRAM, was developed, making it feasible to perform systematic and quantitative analyses. The STRAM considers both geological risk factors and TBM types, such as EPB open mode, EPB closed mode, and slurry TBM, and it suggests the potential risk events that may occur during tunnel construction. The degree of risk is quantified by estimating the cost of counter-measures. A risk management methodology, based on the STRAM and to be used in the design stage of a shield TBM project, is proposed in order to obtain the optimized TBM type for the specific conditions. With the methodology, it is also possible to identify potential risk events that can occur even with the best machine type, and thus, the tunnel designer can optimize the details of the selected machine during design stage. The proposed methodology was applied to an undersea tunnel project to verify its validity. The various risk events were identified based on geological risk factors for that project, and the slurry TBM was selected as the best machine type after analyzing the potential risk events in a quantitative manner. Generally, a slurry TBM is used for undersea tunnel construction to cope with high water pressure, which was confirmed in this case study by using the proposed risk analysis model. ACKNOWLEDGMENTS This research was supported by a grant (Project number: 17SCIP-B066321-05, Development of Key Subsea Tunneling Technology) from the Infrastructure and Transportation Technology Promotion Research Program (Project number: 17RTRP-B104237-03, Development of Railway Infrastructure Management Standard Technology based on BIM) from the Railway Technology Research Program funded by the Ministry of Land, Infrastructure and Transport of the Korean government.

REFERENCES Benardos, A.G. and Kaliampakos, D.C. (2004), “A methodology for assessing

geotechnical hazards for TBM tunnelling-illustrated by the Athens Metro, Greece”, International Journal of Rock Mechanics and Mining Sciences, 41, 987-999.

Chong, W. (2013), “Tunnel Boring Machine (TBM) performance in Singapore’s Mass Rapid Transit (MRT) system”, Master Thesis, Massachusetts Institute of Technology, Cambridge, MA, USA.

Jensen, F.V. (2001), Bayesian Networks and Decision Graphs, Springer-Verlag, New York, NY, USA.

Koh, S.Y., Kwon, S.J., Choo, S.Y. and Kim, Y.M. (2010), “The study of the disputed issues during the soft ground shield TBM design and construction according to shield TBM trouble case study”, 2010 Autumn Conference of the Korean Society for Railway, Jeju, Korea, 2362-2371 (in Korean).

Kwak, J.H. and Park, H.K. (2009), “A case study of delay analysis for E.P.B shield TBM method in construction site”, Journal of the Korean Society of Civil Engineers, KSCE, 29(6D), 737-743 (in Korean).

Park, J. (2015), “A risk management system applicable to shield TBM tunnel using Bayesian network”, Ph.D. Dissertation, Korea University, Seoul, Korea (in Korean).

Park, J., Chung, H., Moon, J.B., Choi, H. and Lee, I.M. (2016), “Overall risk analysis of shield TBM tunnelling using Bayesian Networks (BN) and Analytic Hierarchy Process (AHP)”, Journal of Korean Tunnelling and Underground Space Association, 18(5), 453-467 (in Korean).

Parsons (2008), Risk Register. Shirlaw, J.N., Hencher, S.R. and Zhao, J. (2000), “Design and construction issues for

excavation and tunnelling in some tropically weathered rocks and soils”, Proceedings of GeoEng2000, Melbourne, Australia, 1, 1286-1329.

Sousa, R.L. and Einstein, H.H. (2012), “Risk analysis during tunnel construction using Bayesian Networks: Porto Metro case study”, Tunnelling and Underground Space Technology, 27, 86-100.

Suh, Y.H., Nam, H.L., Chang, S., Lee, J.W., Lee, S.C., Kwon, Y.W. and Jeon, Y.K. (2010), “Geotechnical investigation and risk analysis of the first subsea tunnel by conventional tunneling method in Korea”, 2010 Conference of the Korean Society for Rock mechanics, Daejeon, Korea, 101-109 (in Korean).

Tóth, Á ., Gong, Q. and Zhao, J. (2013), "Case studies of TBM tunneling performance in rock-soil interface mixed ground”, Tunnelling and Underground Space Technology, 38, 140-150.